regex-vs-llm-structured-text

Hybrid regex-and-LLM framework for parsing structured text, optimizing cost by handling 95–98% with regex and reserving LLM calls for edge cases. A practical decision framework for parsing structured text (quizzes, forms, invoices, documents). The key insight: regex handles 95-98% of cases cheaply and deterministically. Reserve expensive LLM calls for the remaining edge cases. Flag items that may need LLM review: From a production quiz parsing pipeline (410 items): - Combines regex extraction with confidence scoring to flag low-confidence items, then validates only those items with an LLM, reducing LLM calls by ~95% versus all-LLM approaches - Includes production-ready Python patterns for regex parsing, confidence scoring, and hybrid pipeline orchestration with real metrics from a 410-item quiz parsing example - Best suited for structured, repeating text patterns like quizzes, forms, invoices, and documents where deterministic extraction is possible - Emphasizes test-driven development, immutable data structures, and metric logging to track pipeline health and identify when regex thresholds degrade - Parsing structured text with repeating patterns (questions, forms, tables) - Deciding between regex and LLM for text extraction - Building hybrid pipelines that combine both approaches - Optimizing cost/accuracy tradeoffs in text processing - Start with regex — even imperfect...